A system and method for detecting foreign matter content in metal powder

Abstract

The present invention relates to the technical field of metal physical property analysis, and in particular to a system and method for detecting impurity content in metal powder. The method comprises the steps of: obtaining each superpixel block in a grayscale image of metal powder, wherein the superpixel block contains a plurality of pixel points; performing convex hull detection on each superpixel block to obtain the number of convex edges of the superpixel block, wherein the convex edge contains a plurality of pixel points; determining the curvature of the pixel points in the convex edge of the superpixel block by a three-point method, taking the difference between the curvature of each pixel point and the mean curvature of the pixel points in the convex edge where the pixel point is located as the first difference, determining the arc of the convex edge where the pixel point is located; determining the circularity of the convex edge in the superpixel block; calculating the impurity possibility of the superpixel block; obtaining the impurity content in the metal powder by comparing the impurity possibility of the superpixel block with a threshold value, so that the accuracy of the impurity content in the obtained metal powder can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of metal physical property analysis, and in particular to a system and method for detecting impurity content in metal powder. Background Art

[0002] Metal powder is a key raw material in the powder metallurgy industry. Through processes such as pressing and sintering, metal products with specific shapes, sizes and properties, such as gears and bearings, can be prepared. In addition, metal powder also plays an important role in processes such as metal powder bed melting, binder jetting, thermal spraying, and plasma spraying. Therefore, in order to improve product performance and reduce the scrap rate in the production process, the prepared metal powder needs to be tested for impurities before use.

[0003] The existing patent application document with publication number JP2010127702A discloses an automatic detection method of metal powder foreign matter in an insulating resin composition, which detects metal powder as foreign matter in the insulating resin composition by changing the X-ray irradiation angle using an X-ray fluoroscopy device having at least one X-ray tube. The image of the metal powder is automatically recognized by using an image processing device.

[0004] Although the above scheme can detect and identify metal powder foreign matter with a maximum length of more than 50 μm in the insulating resin, it has high requirements for the object to be processed and does not record the specific steps of how to adjust the X-ray irradiation angle to obtain foreign matter detection results.

[0005] Based on this, how to accurately obtain the detection results of the impurity content in metal powder is a technical problem that needs to be solved urgently by technical personnel in this field. Summary of the invention

[0006] In order to solve the above-mentioned technical problem of how to accurately obtain the detection result of the impurity content in the metal powder, the present invention provides a system and method for detecting the impurity content in the metal powder.

[0007] In a first aspect, the present invention provides a method for detecting the content of impurities in metal powder, which adopts the following technical solution:

[0008] A method for detecting impurity content in metal powder comprises the steps of:

[0009] Obtain each superpixel block in the grayscale image of the metal powder, wherein the superpixel block contains a plurality of pixel points; perform convex hull detection on each superpixel block to obtain the number of convex edges of the superpixel block, wherein the convex edge contains a plurality of pixel points; determine the curvature of the pixel points in the convex edge of the superpixel block by a three-point method, take the difference between the curvature of each pixel point and the mean curvature of the pixel points in the convex edge where the pixel point is located as a first difference, determine the arc of the convex edge where the pixel point is located, and the arc is negatively correlated with the absolute value of the first difference; determine the circularity of the convex edge in the superpixel block, and the circularity is positively correlated with the arc; calculate the impurity possibility of the superpixel block, and the impurity possibility is negatively correlated with the circularity of the convex edge of the superpixel block; obtain the impurity content in the metal powder by comparing the impurity possibility of the superpixel block with a threshold.

[0010] The present invention takes into account the significant difference in physical shapes between metal powder and impurities. Therefore, the degree of circularity of each super-pixel block in the grayscale image of the metal powder is obtained to obtain the impurity possibility of the super-pixel block. The higher the degree of circularity, the lower the impurity possibility. In this way, the impurity content in the metal powder can be accurately obtained.

[0011] According to a method for detecting impurity content in metal powder provided by the present invention, the acquisition of each superpixel block in the metal powder grayscale image includes: using an electron microscope to acquire a microscopic image of the metal powder, and graying the microscopic image to obtain a metal powder grayscale image; using a superpixel segmentation algorithm to segment the metal powder grayscale image and remove the background area to obtain multiple superpixel blocks.

[0012] The present invention takes into account that the background area in the metal powder grayscale image will be merged with the debris area with a lower grayscale value during superpixel segmentation, so the influence of the background area on the debris content detection is reduced by removing the background area.

[0013] According to a method for detecting impurity content in metal powder provided by the present invention, convex hull detection is performed on each superpixel block to obtain the number of convex edges of the superpixel block, including: presetting an empirical threshold for edge concave detection; obtaining corner points on the edge of the superpixel block by a corner point detection algorithm, and using a convex hull algorithm to obtain the convex hull of the superpixel block corner points and the total area of ​​the convex hull; obtaining convex hull edge segments corresponding to adjacent corner points in the convex hull, and determining the area of ​​a closed area enclosed by the edge between adjacent corner points on the superpixel block and the convex hull edge segments corresponding to the adjacent corner points; obtaining the ratio of the closed area area to the total area of ​​the convex hull, if the ratio is less than or equal to the empirical threshold for edge concave detection, the edge between adjacent corner points on the superpixel block is a convex edge, otherwise it is a concave edge; and finally obtaining the number of convex edges in the superpixel block.

[0014] According to a method for detecting impurity content in metal powder provided by the present invention, the arc angle of the convex edge satisfies the relationship:

[0015] ;

[0016] In the formula, Indicates The first superpixel block The radius of the convex edge, Indicates The first superpixel block The number of pixels contained in a convex edge, Indicates The first superpixel block The convex edge The curvature of a pixel, represents the linear normalization function, Represents the absolute value symbol.

[0017] The present invention provides an accurate method for calculating the circular arc of a convex edge. The circular arc of the convex edge can be accurately obtained by characterizing the bending degree at a pixel point by curvature.

[0018] According to a method for detecting impurity content in metal powder provided by the present invention, the determining of the degree of roundness of the convex edge in the superpixel block comprises: taking the difference between the mean curvature of the pixel points in the convex edge and the mean curvature of the pixel points in all convex edges in the superpixel block as a second difference, determining the curvature difference of the convex edge in the superpixel block, wherein the curvature difference is positively correlated with the absolute value of the second difference; and calculating the degree of roundness of the convex edge in the superpixel block:

[0019] ;

[0020] In the formula, Indicates The circularity of the superpixel block, Indicates The number of convex edges of a superpixel block, Indicates The first superpixel block The radius of the convex edge, Indicates The first superpixel block The difference in curvature of the convex edges.

[0021] According to a method for detecting impurity content in metal powder provided by the present invention, the curvature difference of the convex edge satisfies the relationship:

[0022] ;

[0023] In the formula, Indicates The first superpixel block The curvature difference of the convex edges, Indicates The first superpixel block The number of pixels contained in a convex edge, Indicates The first superpixel block The convex edge The curvature of a pixel, Indicates The number of convex edges of a superpixel block, represents the linear normalization function, Represents the absolute value symbol.

[0024] According to a method for detecting impurity content in metal powder provided by the present invention, the calculating of the impurity possibility of the super pixel block comprises: determining the circularity distinction credibility of the super pixel block, wherein the circularity distinction credibility is positively correlated with the number of pixels in the convex edge; calculating the credible circularity of the super pixel block:

[0025] ;

[0026] In the formula, Indicates The degree of credible circularity of superpixel blocks, Indicates The circularity of each superpixel block is used to distinguish the credibility. Indicates The circularity of a super pixel block is determined; and the probability of debris of the super pixel block is determined, wherein the probability of debris is negatively correlated with the credible circularity.

[0027] The present invention also takes into account that the concave edge of the superpixel block may be the edge of an obstructed metal powder particle or the irregular edge of debris, and the metal powder particle has a higher roundness than the debris in the image. Therefore, by obtaining the proportion of convex edges and correcting the circularity to obtain a credible circularity, the obstructed metal powder particles can be effectively avoided from being misclassified as debris, thereby effectively improving the accuracy of the debris content in the obtained metal powder.

[0028] According to a method for detecting impurity content in metal powder provided by the present invention, the impurity possibility of the super pixel block satisfies the relationship:

[0029] ;

[0030] In the formula, Indicates The probability of clutter in superpixel blocks, Indicates The degree of credible circularity of superpixel blocks, represents the linear normalization function, Indicates The grayscale mean of the superpixel blocks.

[0031] The present invention also takes into account the different grayscale values ​​between metal powder and impurities. Therefore, by combining the circularity and grayscale value of the superpixel block, the impurity possibility of the superpixel block can be accurately obtained, thereby accurately obtaining the impurity content detection result in the metal powder.

[0032] According to a method for detecting impurity content in metal powder provided by the present invention, the impurity content in the metal powder is obtained by comparing the impurity possibility of the superpixel block with a threshold, comprising: presetting a threshold; recording the superpixel block whose impurity possibility is greater than the threshold as an impurity superpixel block, and recording the superpixel block whose impurity possibility is less than or equal to the threshold as a metal powder superpixel block, and obtaining the area of ​​all impurity superpixel blocks and the area of ​​the metal powder superpixel block in the metal powder grayscale image; obtaining the sum of the areas of the impurity superpixel block and the metal powder superpixel block in the metal powder grayscale image, and taking the ratio of the area of ​​the impurity superpixel block to the sum of the areas as the impurity content in the metal powder.

[0033] In a second aspect, the present invention provides a system for detecting the content of impurities in metal powder, which adopts the following technical solution:

[0034] A system for detecting impurity content in metal powder comprises: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned method for detecting impurity content in metal powder is implemented.

[0035] By adopting the above technical solution, the above-mentioned method for detecting the impurity content in a metal powder is generated into a computer program and stored in a memory so as to be loaded and executed by a processor, thereby making a terminal device based on the memory and the processor for easy use.

[0036] The present invention has the following technical effects:

[0037] Based on the above technical scheme, when determining the impurity content in metal powder, the present invention obtains the circularity of each super-pixel block in the grayscale image of the metal powder, thereby obtaining the impurity possibility of the super-pixel block, and can distinguish the metal powder and the impurities based on the physical shape difference, thereby accurately obtaining the impurity content in the metal powder; in addition, the present invention also corrects the circularity by obtaining the convex edge ratio to obtain a credible circularity, and accurately obtains the impurity possibility of the super-pixel block by combining the credible circularity and grayscale value of the super-pixel block, effectively avoiding the situation where obscured metal powder particles are mistakenly classified as impurities, thereby effectively improving the accuracy of the impurity content in the obtained metal powder. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts.

[0039] Figure 1 A schematic flow chart of a method for detecting impurity content in metal powder provided by an embodiment of the present invention;

[0040] Figure 2 A metal powder grayscale image provided by an embodiment of the present invention;

[0041] Figure 3 A schematic diagram of superpixel segmentation results of a metal powder grayscale image provided by an embodiment of the present invention;

[0042] Figure 4 A schematic diagram of a superpixel segmentation result of a metal powder grayscale image after removing the background area provided by an embodiment of the present invention;

[0043] Figure 5 A schematic diagram of the relationship between a superpixel block and a convex hull of a corner point provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0045] It should be understood that when the terms "first", "second", etc. are used in the claims, descriptions, and drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the description and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections.

[0046] It should be noted that the impurity content in the metal powder will affect the use effect of the metal powder. In order to improve the use effect of the metal powder and thus improve the service life and safety of the materials obtained based on the metal powder, the impurity content of the metal powder needs to be tested after the metal powder is mixed.

[0047] Based on this, the embodiment of the present invention discloses a method for detecting the content of impurities in metal powder. Figure 1 As shown, Figure 1 A schematic flow chart of a method for detecting impurity content in metal powder provided in an embodiment of the present invention, the method comprising the following steps S1 to S6:

[0048] S1: Obtain each super pixel block in the grayscale image of the metal powder, where the super pixel block contains multiple pixel points.

[0049] It should be noted that when the impurity content of the metal powder is measured after the batching is completed, it can be obtained by collecting and analyzing the microscopic image of the metal powder. This not only saves storage space, but also facilitates the subsequent analysis of the internal area characteristics of the image through the gray value characteristics of the image, so as to accurately obtain the impurity content in the metal powder. In the process of using metal powder, in order to improve the coating quality and ensure that the powder can be heated and melted more evenly, the metal powder is generally spherical or nearly spherical in shape, and the shape in the obtained microscopic image is circular or nearly circular, while the impurities in the metal powder tend to be irregular in shape. By analyzing the difference in the physical appearance of the metal powder and the impurities, the metal powder and the impurities can be distinguished.

[0050] For example, in an embodiment of the present invention, when obtaining each superpixel block in a metal powder grayscale image, an electron microscope can be used to obtain a microscopic image of the metal powder, and the microscopic image can be grayscaled to obtain a metal powder grayscale image; a superpixel segmentation algorithm can be used to segment the metal powder grayscale image and remove the background area to obtain multiple superpixel blocks.

[0051] For example, the grayscale image of metal powder obtained after grayscale processing can be seen in Figure 2 As shown, Figure 2 A metal powder grayscale image provided by an embodiment of the present invention.

[0052] It should be noted that the superpixel segmentation algorithm usually retains the important edges and details in the image better. Therefore, the superpixel segmentation algorithm can be used to identify the shape and boundaries of metal powder particles and debris in the microscopic image. In addition, since the superpixel block segmentation result divides the actual particle area or debris area into multiple adjacent superpixel blocks, and the grayscale mean difference in the multiple superpixel blocks constituting the actual particle area or debris area is small, the adjacent superpixel blocks with small grayscale mean difference in the area can be merged.

[0053] For example, when the metal powder grayscale image is segmented by the superpixel segmentation algorithm, the metal powder grayscale image can be first divided into initial superpixel block regions by the superpixel segmentation algorithm, the grayscale mean value in the initial superpixel block region is obtained, the merging threshold T=25 is set, all initial superpixel blocks are traversed, and adjacent superpixel blocks whose absolute difference in the grayscale mean value in the region is less than the merging empirical threshold are merged to obtain the final superpixel segmentation result. For details, please refer to Figure 3 As shown, Figure 3 A schematic diagram of superpixel segmentation results of a metal powder grayscale image provided by an embodiment of the present invention. Figure 3 It can be seen that the metal powder grayscale image can be divided into multiple superpixel blocks through the superpixel segmentation algorithm.

[0054] The merging threshold can be set according to actual needs.

[0055] In addition, the embodiment of the present invention takes into account that the background area will be merged with the debris area with lower grayscale values, and the background area itself will form a superpixel block. Therefore, in order to effectively identify the debris in the metal powder, it is also necessary to remove the background area in the metal powder microscopic image to avoid the influence of the background area on the recognition result.

[0056] Specifically, the metal powder grayscale image is divided into a white mask and a black mask by threshold segmentation, see Figure 4 As shown, Figure 4 A schematic diagram of the superpixel segmentation result of a metal powder grayscale image after removing the background area provided by an embodiment of the present invention, wherein the white mask is the area with a grayscale value of 255, and the black mask is the area with a grayscale value of 0. The number of empirical screening L is set to 3, and the black mask area whose number of pixels in the connected domain is greater than the number of empirical screening is recorded as the background area. The number of empirical screening can be set according to actual needs.

[0057] For example, in order to make the image clearer, the image may be denoised by Gaussian filtering; the denoising method may be specifically set according to actual needs, and the embodiment of the present invention is not limited thereto.

[0058] After the microscopic image of the metal powder is processed based on the above method to obtain multiple super pixel blocks, the super pixel blocks can be further processed based on the following steps.

[0059] S2: Perform convex hull detection on each superpixel block to obtain the number of convex edges of the superpixel block.

[0060] It should be noted that when metal powder particles overlap, the intersection of the overlapping edges of the particles in the microscopic image appears as a sharp corner point on the obscured edge of the particle, and the larger the area formed by the edge line between the two nearest adjacent corner points that can be connected along the edge and the line between the two points in the convex hull detection result, the more likely the edge is to be concave.

[0061] Since there are also depressions in the edges of debris, the concave edges may be caused by metal powder particles blocking or the edges of debris. Based on the above steps, it can be known that the shape of metal powder particles is circular or nearly circular, and debris is generally irregular. Therefore, the arc of metal powder particles is higher than that of debris. If the arc of the convex edge is higher, the possibility that the superpixel block where the convex edge is located is metal powder particles is higher, and the possibility that it is debris is lower. Based on this, the convex edges in each superpixel block can be obtained first.

[0062] By way of example, in an embodiment of the present invention, convex hull detection is performed on each superpixel block to obtain the number of convex edges of the superpixel block, and an empirical threshold for edge concave detection can be preset; the corner points on the edge of the superpixel block are obtained by a corner point detection algorithm, and the convex hull of the corner points of the superpixel block and the total area of ​​the convex hull are obtained by a convex hull algorithm; the convex hull edge segments corresponding to adjacent corner points are obtained in the convex hull, and the area of ​​the closed area enclosed by the edges between adjacent corner points on the superpixel block and the convex hull edge segments corresponding to the adjacent corner points are determined; the ratio of the area of ​​the closed area to the total area of ​​the convex hull is obtained, and if the ratio is less than or equal to the empirical threshold for edge concave detection, the edge between adjacent corner points on the superpixel block is a convex edge, otherwise it is a concave edge; finally, the number of concave edges and the number of convex edges in the superpixel block are obtained.

[0063] The concave edge and convex edge in the superpixel block finally obtained contain multiple pixels. The edge concave detection empirical threshold can be preset as . The empirical threshold for edge depression detection can be set according to actual needs. The corner detection algorithm can be Shi-Tomasi corner detection, Harris corner detection, etc., which can be set according to actual needs. Using a convex hull algorithm to obtain the convex hull of the corner points of a superpixel block and the total area of ​​the convex hull is a prior art, and the embodiments of the present invention will not be described in detail here.

[0064] The edge between adjacent corner points detected in the superpixel block is the edge of the superpixel block; the convex hull edge is divided into convex hull edge segments by the corner points.

[0065] For example, see Figure 5 As shown, Figure 5 A schematic diagram of the relationship between a superpixel block and a convex hull of a corner point provided by an embodiment of the present invention. In which, A1, A2, A3, A4 and A5 represent corner points, a circle represents a superpixel block, the dotted line portion surrounded by A1, A2, A3, A4 and A5 is the convex hull edge of the corner point, the oblique line portion represents the closed area area surrounded by the edge between adjacent corner points and the convex hull edge segments corresponding to the adjacent corner points, S1 represents the closed area area surrounded by the edge between adjacent corner points A4 and A5 and the convex hull edge segments corresponding to A4 and A5; S2 represents the closed area area surrounded by the edge between adjacent corner points A1 and A2 and the convex hull edge segments corresponding to A1 and A2; S3 represents the closed area area surrounded by the edge between adjacent corner points A2 and A3 and the convex hull edge segments corresponding to A2 and A3.

[0066] After processing the above steps, we can get Figure 5 The edges between adjacent corner points A1 and A2, A2 and A3, A4 and A5 are concave edges, and the edges between adjacent corner points A1 and A5, A3 and A4 are convex edges. At the same time, the edges between A1 and A5, A3 and A4 are also the edges of the superpixel blocks of the blocked metal powder particles. Finally, the number of convex edges of each superpixel block in the metal powder grayscale image is obtained, and the following steps are continued.

[0067] S3: Determine the curvature of the pixel points in the convex edge of the superpixel block by a three-point method, take the difference between the curvature of each pixel point and the average curvature of the pixel points in the convex edge where the pixel point is located as the first difference, and determine the arc of the convex edge where the pixel point is located.

[0068] It should be noted that since there are also depressions in the edges of debris, the concave edges of the superpixel blocks may be the edges of obscured metal powder particles or the irregular edges of debris. The metal powder particles have a higher roundness than the debris in the image. The higher the roundness of the convex edge, the more likely it is that the superpixel block area is a metal powder particle. Therefore, when the proportion of convex edges is higher, the credibility of judging whether the superpixel block area is debris based on the degree of circularity is higher.

[0069] Based on this, when determining the probability of debris in a superpixel block, it is necessary to first calculate the circularity distinction credibility of the superpixel block.

[0070] Among them, determining the curvature of the pixel points in the convex edge of the super pixel block by the three-point method is a prior art, and the embodiment of the present invention will not be described in detail here.

[0071] Specifically, the edge of the superpixel block includes a concave edge and / or a convex edge. Since the edge of the superpixel block is a one-dimensional linear structure, the number of pixels contained in the edge is the length of the edge. When the length of the convex edge in the superpixel block is high, the credibility of distinguishing the superpixel block by the degree of circularity is higher. To determine the credibility of distinguishing the degree of circularity of the superpixel block, please refer to the following relationship:

[0072] ;

[0073] In the formula, Indicates The circularity of each superpixel block is used to distinguish the credibility. Indicates The number of convex edges of a superpixel block, Indicates The first superpixel block The number of pixels contained in a convex edge, Indicates The number of concave edges of a superpixel block, Indicates The number of pixels contained in the i-th concave edge of a superpixel block.

[0074] In the above formula, Indicates The length of the concave edge in the superpixel block, Indicates The length of the convex edge in a superpixel block. The higher the proportion of the convex edge length, the higher the credibility of the circularity of the superpixel block.

[0075] Based on the above steps, the convex edges of each superpixel block in the metal powder grayscale image can be obtained. When the overall arc degree of all convex edges in the superpixel block is higher and the overall curvature of each convex edge is more similar, the figure formed by all convex edges conforms more to the standard circle, and the higher the circular degree of the superpixel block area to which the convex edge belongs, the higher the possibility that the superpixel block is a metal powder particle and the lower the possibility that it is a debris. The curvature can be used to characterize the curvature of the curve, based on this, the arc degree of the convex edge can be obtained based on the curvature of each pixel point in the convex edge.

[0076] Specifically, in the embodiment of the present invention, the arc degree of the convex edge is determined, and the following relationship can be referred to for details:

[0077] ;

[0078] In the formula, Indicates The first superpixel block The radius of the convex edge, Indicates The first superpixel block The number of pixels contained in a convex edge, Indicates The first superpixel block The convex edge The curvature of a pixel, represents the linear normalization function, Represents the absolute value symbol.

[0079] In the above formula, Indicates The first superpixel block The mean curvature of the pixels in the convex edge, represents the first difference, The first superpixel block The curvature of each pixel in the convex edge is The larger the difference in the mean curvature of the pixels in the first convex edge, the The first superpixel block The more unstable the curvature of a convex edge is, the lower the possibility of it being an arc, and the arc degree of the convex edge is negatively correlated with the absolute value of the first difference.

[0080] After the arc degree of the convex edge is obtained based on the above steps, the circular degree of the convex edge can be obtained based on the arc degree of the convex edge.

[0081] S4: Determine the degree of roundness of convex edges in the superpixel block.

[0082] It should be noted that if the curvature at the pixel point in the convex edge of the superpixel block shows a large fluctuation, it means that the shape of the convex edge is not a smooth arc, but may contain multiple curved segments or inflection points. The possibility of the convex edge being an arc is low, and the corresponding circularity is also lower; in addition, if the mean curvature of one of the convex edges of the superpixel block is significantly different from the mean curvature of other convex edges, it means that the convex edge is less similar to other convex edges, and the superpixel block is less close to a standard circle. Based on this, the circularity of the convex edge in the superpixel block can be obtained by the difference in the mean curvature between the convex edges.

[0083] By way of example, in an embodiment of the present invention, when determining the degree of roundness of a convex edge in a superpixel block, the difference between the mean curvature of a pixel point and the mean curvature of the pixel points in all convex edges of the superpixel block where the pixel point is located can be used as a second difference to determine the curvature difference of the convex edge in the superpixel block, and the curvature difference is positively correlated with the absolute value of the second difference; the degree of roundness of the convex edge in the superpixel block is calculated based on the circularity and curvature difference of the convex edge, and the degree of roundness is positively correlated with the circularity.

[0084] For example, in the embodiment of the present invention, the curvature difference of the convex edge is determined, and the following relationship can be specifically referred to:

[0085] ;

[0086] In the formula, Indicates The first superpixel block The curvature difference of the convex edges, Indicates The first superpixel block The number of pixels contained in a convex edge, Indicates The first superpixel block The convex edge The curvature of a pixel, Indicates The number of convex edges of a superpixel block, represents the linear normalization function, Represents the absolute value symbol.

[0087] In the above formula, Indicates The mean curvature of all convex edges of a superpixel block, Indicates The first superpixel block The greater the difference between the two, the greater the curvature of the convex edge. The first superpixel block The convex edge and The larger the curvature difference of other convex edges in the superpixel block, the The less the superpixel block is close to the standard circle.

[0088] After respectively obtaining the arc degree and curvature difference of the convex edge based on the above steps, the circularity of the convex edge in the superpixel block can be calculated based on the arc degree and curvature difference of the convex edge.

[0089] Specifically, the degree of roundness of the convex edge can be determined by referring to the following formula:

[0090] ;

[0091] In the formula, Indicates The circularity of the superpixel block, Indicates The number of convex edges of a superpixel block, Indicates The first superpixel block The radius of the convex edge, Indicates The first superpixel block The difference in curvature of the convex edges.

[0092] In the above formula, Indicates The first superpixel block The possibility that the convex edge is a standard arc, the larger the value, the higher the probability that the convex edge is a standard arc. The first superpixel block The curvature fluctuation of the pixel points in the convex edge is small, the arc degree is high, and the curvature similarity with other convex edges is high. Correspondingly, The higher the probability that the convex edge in the superpixel block is a standard arc, the The higher the circularity of the superpixel block, the better.

[0093] After obtaining the circularity of the superpixel block based on the above steps, the possibility of debris in the superpixel block can be obtained based on the circularity of the superpixel block. The higher the circularity of the superpixel block, the lower the possibility of debris in the superpixel block, that is, perform the following steps.

[0094] S5: Calculate the probability of clutter for the superpixel block.

[0095] It should be noted that, based on the above steps, the circularity distinction credibility of the superpixel block can be obtained. When the circularity distinction credibility is high, metal powder particles and debris can be directly distinguished; however, when the circularity distinction credibility is low, it is impossible to directly determine whether the superpixel block area has overlapping metal powder particle areas or irregular edges of debris. At this time, if the circularity is directly calculated, the overlapping metal powder particles will be mistaken for debris, resulting in the final accuracy of the debris content detection result being low. For such situations, the embodiment of the present invention takes into account that even if the circularity distinction credibility is low, the convex edge of the debris still does not have a high circularity of the convex edge of the overlapping obscured metal powder particle. Based on this, the embodiment of the present invention handles such situations by calculating the credible circularity. When the circularity distinction credibility is low, the circularity value should be closer to 0.5 to reduce the situation where overlapping metal powder particles are mistaken for debris.

[0096] By way of example, in an embodiment of the present invention, calculating the debris possibility of a superpixel block includes: determining the circularity distinction credibility of the superpixel block, the circularity distinction credibility is positively correlated with the number of pixels in the convex edge; calculating the credible circularity of the superpixel block, determining the debris possibility of the superpixel block, the debris possibility of the superpixel block is negatively correlated with the credible circularity of the superpixel block, and the debris possibility is negatively correlated with the circularity of the convex edge of the superpixel block.

[0097] Specifically, in the embodiment of the present invention, the degree of credible circularity of the superpixel block is determined by referring to the following relationship:

[0098] ;

[0099] In the formula, Indicates The degree of credible circularity of superpixel blocks, Indicates The circularity of each superpixel block is used to distinguish the credibility. Indicates The circularity of a superpixel.

[0100] In the above formula, Indicates the contribution of the concave edge to the credible circularity. When the circularity distinction credibility is low, the concave edge length accounts for a high proportion, and it is not possible to effectively distinguish between metal powder particles and debris. In this case, the circularity needs to be directly set to 0.5. It represents the contribution of the convex edge to the credible degree of circularity. When the circularity distinction credibility is high, the proportion of the convex edge length is high, and the credibility of the circularity obtained by the arc of the convex edge is higher.

[0101] Based on the above formula, the credible circularity of the superpixel block can be obtained by the ratio of the concave edge length to the convex edge length in the superpixel block, and thus the possibility of debris in the superpixel block can be obtained based on the credible circularity of the superpixel block.

[0102] Specifically, in the embodiment of the present invention, the metal powder particles and the debris have both physical shape differences and grayscale value differences. The overall grayscale value of the debris is low, and the overall grayscale value of the metal powder particles is high. Based on this, the grayscale value and the credible circularity of the superpixel block can be combined to jointly determine the possibility of debris in the superpixel block. For details, see the following relationship:

[0103] ;

[0104] In the formula, Indicates The probability of clutter in superpixel blocks, Indicates The degree of credible circularity of superpixel blocks, represents the linear normalization function, Indicates The grayscale mean of the superpixel blocks.

[0105] In the above formula, Indicates The probability that the superpixel block is a metal powder particle is high. When the credible circularity of the superpixel block is high and the grayscale mean is high, the superpixel block is closer to a circle and the overall color is white, which is consistent with the characteristics of metal powder particles in the microscopic image. The lower the possibility that a super-pixel block is a metal powder particle, the higher the possibility that the super-pixel block is a foreign object.

[0106] Based on the above steps, the impurity probability of each super pixel block can be obtained, and the impurity content in the metal powder can be obtained by comparing the impurity probability of the super pixel block with the threshold.

[0107] S6: Obtain the impurity content in the metal powder by comparing the impurity possibility of the super pixel block with the threshold.

[0108] It should be noted that, in the grayscale image of the metal powder after removing the background area, the sum of the debris area and the metal powder area is the total area. The higher the proportion of the debris area in the total area, the higher the debris content.

[0109] By way of example, in an embodiment of the present invention, when obtaining the impurity content in the metal powder by comparing the impurity possibility of the superpixel block with the threshold, a threshold can be preset; the superpixel block with an impurity possibility greater than the threshold is recorded as a impurity superpixel block, and the superpixel block with an impurity possibility less than or equal to the threshold is recorded as a metal powder superpixel block, and the area of ​​all the impurity superpixel blocks and the area of ​​the metal powder superpixel blocks in the metal powder grayscale image are obtained; the sum of the areas of the impurity superpixel blocks and the metal powder superpixel blocks in the metal powder grayscale image is obtained, and the ratio of the area of ​​the impurity superpixel block to the sum of the areas is used as the impurity content in the metal powder.

[0110] The threshold value may be preset to 0.8, and may be set specifically according to actual needs.

[0111] After obtaining the impurity content in the metal powder based on the above embodiment, a threshold value of the impurity content can also be set. When the impurity content in the current metal powder is detected multiple times, if the average impurity content in the metal powder exceeds the threshold value of the impurity content, it means that the quality of the current metal powder is unqualified and needs to be reworked.

[0112] The debris content threshold can be set to 0.1, and can be set specifically according to actual needs.

[0113] It can be seen that in the embodiment of the present invention, when obtaining the impurity content in the metal powder, each superpixel block in the grayscale image of the metal powder can be obtained, and the superpixel block contains multiple pixel points; convex hull detection is performed on each superpixel block to obtain the number of convex edges of the superpixel block, and the convex edge contains multiple pixel points; the curvature of the pixel points in the convex edge of the superpixel block is determined by the three-point method, and the difference between the curvature of each pixel point and the mean curvature of the pixel points in the convex edge where the pixel point is located is taken as the first difference, and the arc of the convex edge where the pixel point is located is determined, and the arc is negatively correlated with the absolute value of the first difference; the circularity of the convex edge in the superpixel block is determined, and the circularity is positively correlated with the arc; the impurity possibility of the superpixel block is calculated, and the impurity possibility is negatively correlated with the circularity of the convex edge of the superpixel block; the impurity content in the metal powder is obtained by comparing the impurity possibility of the superpixel block with the threshold.

[0114] Thus, the embodiment of the present invention takes into account the great difference between the physical shapes of metal powder and debris, and therefore obtains the degree of circularity of each super-pixel block in the grayscale image of the metal powder, thereby obtaining the debris possibility of the super-pixel block, and the higher the degree of circularity, the lower the debris possibility; in addition, the embodiment of the present invention also takes into account that the concave edge of the super-pixel block may be the edge of the obscured metal powder particle or the irregular edge of the debris, and the circularity of the metal powder particle in the image is higher than that of the debris, and therefore, the circularity is corrected by obtaining the proportion of convex edges to obtain a credible circularity, and the credible circularity and grayscale value of the super-pixel block are combined to accurately obtain the debris possibility of the super-pixel block, which can effectively improve the accuracy of the debris content in the obtained metal powder.

[0115] An embodiment of the present invention further discloses a system for detecting impurity content in metal powder, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a method for detecting impurity content in metal powder provided by the present invention is implemented.

[0116] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.

[0117] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM, a dynamic random access memory DRAM, a static random access memory SRAM, an enhanced dynamic random access memory EDRAM, a high bandwidth memory HBM, a hybrid memory cube HMC, etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device.

[0118] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

[0119] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting impurity content in metal powder, characterized in that: include: Acquire each super pixel block in the metal powder grayscale image, wherein the super pixel block includes a plurality of pixel points; Performing convex hull detection on each superpixel block to obtain the number of convex edges of the superpixel block, wherein the convex edge includes a plurality of pixel points; Determine the curvature of the pixel points in the convex edge of the superpixel block by a three-point method, take the difference between the curvature of each pixel point and the average curvature of the pixel points in the convex edge where the pixel point is located as a first difference, determine the arc of the convex edge where the pixel point is located, and the arc is negatively correlated with the absolute value of the first difference; The arc of the convex edge satisfies the relationship: ； In the formula, Indicates The first superpixel block The radius of the convex edge, Indicates The first superpixel block The number of pixels contained in a convex edge, Indicates The first superpixel block The convex edge The curvature of a pixel, represents the linear normalization function, Indicates the absolute value symbol; Determining a degree of roundness of a convex edge in the superpixel block, wherein the degree of roundness is positively correlated with the arc degree; Calculating the probability of debris of the superpixel block, wherein the probability of debris is negatively correlated with the degree of roundness of the convex edge of the superpixel block; The impurity content in the metal powder is obtained by comparing the impurity possibility of the super pixel block with the threshold.

2. The method for detecting impurity content in metal powder according to claim 1, characterized in that: The step of obtaining each super pixel block in the metal powder grayscale image comprises: An electron microscope is used to obtain a microscopic image of the metal powder, and the microscopic image is gray-scaled to obtain a metal powder grayscale image; a superpixel segmentation algorithm is used to segment the metal powder grayscale image and remove the background area to obtain multiple superpixel blocks.

3. The method for detecting impurity content in metal powder according to claim 1, characterized in that: The performing convex hull detection on each superpixel block to obtain the number of convex edges of the superpixel block includes: Preset an empirical threshold for edge concave detection; obtain corner points on the edge of a superpixel block by a corner point detection algorithm, and use a convex hull algorithm to obtain the convex hull of the superpixel block corner points and the total area of ​​the convex hull; obtain the convex hull edge segments corresponding to adjacent corner points in the convex hull, and determine the area of ​​a closed area enclosed by the edge between adjacent corner points on the superpixel block and the convex hull edge segments corresponding to the adjacent corner points; The ratio of the area of ​​the closed region to the total area of ​​the convex hull is obtained. If the ratio is less than or equal to the empirical threshold for edge concave detection, the edge between adjacent corner points on the superpixel block is a convex edge, otherwise it is a concave edge; finally, the number of convex edges in the superpixel block is obtained.

4. The method for detecting impurity content in metal powder according to claim 1, characterized in that: Determining the degree of roundness of the convex edge in the superpixel block comprises: Taking the difference between the mean curvature of the pixel points in the convex edge and the mean curvature of the pixel points in all convex edges in the superpixel block as a second difference, determining the curvature difference of the convex edge in the superpixel block, wherein the curvature difference is positively correlated with the absolute value of the second difference; Calculate the degree of roundness of the convex edges in the superpixel block: ； In the formula, Indicates The circularity of the superpixel block, Indicates The number of convex edges of a superpixel block, Indicates The first superpixel block The radius of the convex edge, Indicates The first superpixel block The difference in curvature of the convex edges.

5. The method for detecting impurity content in metal powder according to claim 4, characterized in that: The curvature difference of the convex edge satisfies the relationship: ； In the formula, Indicates The first superpixel block The curvature difference of the convex edges, Indicates The first superpixel block The number of pixels contained in a convex edge, Indicates The first superpixel block The convex edge The curvature of a pixel, Indicates The number of convex edges of a superpixel block, represents the linear normalization function, Represents the absolute value symbol.

6. The method for detecting impurity content in metal powder according to claim 1, characterized in that: The calculating the probability of debris in the superpixel block comprises: Determining the circularity distinction reliability of the superpixel block, wherein the circularity distinction reliability is positively correlated with the number of pixel points in the convex edge; Calculate the credible circularity of the superpixel block: ； In the formula, Indicates The degree of credible circularity of superpixel blocks, Indicates The circularity of each superpixel block is used to distinguish the credibility. Indicates The circularity of the superpixel blocks; The probability of a clutter of the superpixel block is determined, wherein the probability of a clutter is negatively correlated with the degree of the credible circularity.

7. The method for detecting impurity content in metal powder according to claim 6, characterized in that: The probability of the debris of the super pixel block satisfies the relationship: ； In the formula, Indicates The probability of clutter in superpixel blocks, Indicates The degree of credible circularity of superpixel blocks, represents the linear normalization function, Indicates The grayscale mean of the superpixel blocks.

8. The method for detecting impurity content in metal powder according to claim 1, characterized in that: The method of obtaining the impurity content in the metal powder by comparing the impurity possibility of the super pixel block with the threshold value includes: A threshold is preset; superpixel blocks whose debris possibility is greater than the threshold are recorded as debris superpixel blocks, and superpixel blocks whose debris possibility is less than or equal to the threshold are recorded as metal powder superpixel blocks, and the areas of all debris superpixel blocks and metal powder superpixel blocks in the metal powder grayscale image are obtained; The sum of the areas of the debris superpixel block and the metal powder superpixel block in the metal powder grayscale image is obtained, and the ratio of the area of ​​the debris superpixel block to the sum of the areas is used as the debris content in the metal powder.

9. A system for detecting impurities in metal powder, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a method for detecting impurity content in a metal powder according to any one of claims 1 to 8 is implemented.

Images